Research map / Label-efficient and robust learning / Domain adaptation and generalization
Generalizable Cross-Modality Medical Image Segmentation via Style Augmentation and Dual Normalization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| augment the single source domain into source-similar and source-dissimilar styles, use dual normalization, and select a path by style | U-Net cross-modality segmentation · BraTS, T2 source → target modalities | average Dice (%): 54.44 vs 35.17 | Helped |
| augment the single source domain into source-similar and source-dissimilar styles, use dual normalization, and select a path by style | U-Net cross-modality segmentation · BraTS, T1CE source → target modalities | average Dice (%): 57.98 vs 42.81 | Helped |
| augment the single source domain into source-similar and source-dissimilar styles, use dual normalization, and select a path by style | U-Net cross-modality segmentation · Cardiac MRI → CT | average Dice (%): 51.70 vs 48.06 | Helped |
| augment the single source domain into source-similar and source-dissimilar styles, use dual normalization, and select a path by style | U-Net cross-modality segmentation · Cardiac CT → MRI | average Dice (%): 32.19 vs 18.33 | Helped |
| augment the single source domain into source-similar and source-dissimilar styles, use dual normalization, and select a path by style | U-Net cross-modality segmentation · Abdominal MRI → CT | average Dice (%): 39.62 vs 39.22 | Helped |
| augment the single source domain into source-similar and source-dissimilar styles, use dual normalization, and select a path by style | U-Net cross-modality segmentation · Abdominal CT → MRI | average Dice (%): 70.22 vs 38.91 | Helped |
| use style-based selection instead of committing to one BN path (ablation) | Dual-normalization U-Net · BraTS, source T2 → target modalities | average Dice (%): 30.64; 30.16 | Helped |
| use style-based selection instead of committing to one BN path (ablation) | Dual-normalization U-Net · BraTS, source T1CE → target modalities | average Dice (%): 62.72; 31.30 | Helped |
| vary the number of Bézier-curve control-point pairs from 1 to 5 (ablation) | Style augmentation · BraTS, Cardiac, and Abdominal Multi-Organ datasets | Dice | No clear effect |
| use the proposed method on cross-center prostate segmentation rather than DeepAll | Cross-center prostate segmentation · Prostate MRI cross-center task | Dice (%): 84.19%; 85.46% | Hurt |
| use the proposed method on cross-center prostate segmentation rather than DoFE | Cross-center prostate segmentation · Prostate MRI cross-center task | Dice (%): 84.19%; 87.48% | Hurt |
Closest papers by meaning
- Learning Generalized Medical Image Segmentation from Decoupled Feature Queries (AAAI 2024 · similarity 0.88)
- Modality-agnostic Domain Generalizable Medical Image Segmentation by Multi-Frequency in Multi-Scale Attention (CVPR 2024 · similarity 0.87)
- One-to-Multiple: A Progressive Style Transfer Unsupervised Domain-Adaptive Framework for Kidney Tumor Segmentation (NeurIPS 2024 · similarity 0.87)
- LangDAug: Langevin Data Augmentation for Multi-Source Domain Generalization in Medical Image Segmentation (ICML 2025 · similarity 0.86)
- MAPSeg: Unified Unsupervised Domain Adaptation for Heterogeneous Medical Image Segmentation Based on 3D Masked Autoencoding and Pseudo-Labeling (CVPR 2024 · similarity 0.85)
- Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.